The idea isn’t that the upscaler magically creates information that wasn’t present in the original recording. If that were true, it would be inventing data rather than reconstructing it.
The argument made by Chord is that a sampled waveform contains enough information to reconstruct the original analogue signal, but the accuracy of that reconstruction depends on the digital filter used. A conventional DAC filter is limited by processing power and therefore uses a relatively short filter. Chord’s upscalers use very large FPGA-based filters with vastly more taps, attempting to better estimate what the waveform looked like between the recorded samples.
In theory, this doesn’t add new information; it attempts to recover timing information that is already implicit in the sampled data but has not been reconstructed as accurately by a shorter filter.
Whether that reconstruction is actually audible is the real debate. Many engineers argue that modern DACs already reconstruct the signal well enough that further improvements are insignificant. Chord and some listeners believe the improvements in transient accuracy, imaging and perceived realism remain audible.
So the claim is not “more information out than went in,” but rather “more accurate reconstruction of the information that was already there.
The theory is perfectly reasonable from a signal-processing perspective. The unanswered question is whether the difference between a very good reconstruction and an even better reconstruction is worth £25,000. That’s where the engineering discussion ends and the subjective listening debate begins.
